Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Impact of Image Resolution on Age Estimation with DeepFace and InsightFace

Shiyar Jamo

TL;DR

This work investigates whether external image resolution influences age-estimation accuracy when using pretrained frameworks DeepFace and InsightFace, despite internal resizing to fixed input sizes. By evaluating seven external resolutions on 1000 high-resolution images from IMDB-Clean, the study measures MAE, SD, and MedAE and compares performance with both frameworks. It finds that the optimal external resolution is 224×224 for both models (MAE of 10.83 for DeepFace and 7.46 for InsightFace), while very low or very high resolutions degrade accuracy, with InsightFace also delivering substantially faster processing times. The results suggest practical guidance for real-time systems: use 224×224 external inputs and prefer InsightFace for better accuracy and efficiency, while recognizing potential generalizability limits due to celeb-centric data biases in IMDB-Clean.

Abstract

Automatic age estimation is widely used for age verification, where input images often vary considerably in resolution. This study evaluates the effect of image resolution on age estimation accuracy using DeepFace and InsightFace. A total of 1000 images from the IMDB-Clean dataset were processed in seven resolutions, resulting in 7000 test samples. Performance was evaluated using Mean Absolute Error (MAE), Standard Deviation (SD), and Median Absolute Error (MedAE). Based on this study, we conclude that input image resolution has a clear and consistent impact on the accuracy of age estimation in both DeepFace and InsightFace. Both frameworks achieve optimal performance at 224x224 pixels, with an MAE of 10.83 years (DeepFace) and 7.46 years (InsightFace). At low resolutions, MAE increases substantially, while very high resolutions also degrade accuracy. InsightFace is consistently faster than DeepFace across all resolutions.

Impact of Image Resolution on Age Estimation with DeepFace and InsightFace

TL;DR

This work investigates whether external image resolution influences age-estimation accuracy when using pretrained frameworks DeepFace and InsightFace, despite internal resizing to fixed input sizes. By evaluating seven external resolutions on 1000 high-resolution images from IMDB-Clean, the study measures MAE, SD, and MedAE and compares performance with both frameworks. It finds that the optimal external resolution is 224×224 for both models (MAE of 10.83 for DeepFace and 7.46 for InsightFace), while very low or very high resolutions degrade accuracy, with InsightFace also delivering substantially faster processing times. The results suggest practical guidance for real-time systems: use 224×224 external inputs and prefer InsightFace for better accuracy and efficiency, while recognizing potential generalizability limits due to celeb-centric data biases in IMDB-Clean.

Abstract

Automatic age estimation is widely used for age verification, where input images often vary considerably in resolution. This study evaluates the effect of image resolution on age estimation accuracy using DeepFace and InsightFace. A total of 1000 images from the IMDB-Clean dataset were processed in seven resolutions, resulting in 7000 test samples. Performance was evaluated using Mean Absolute Error (MAE), Standard Deviation (SD), and Median Absolute Error (MedAE). Based on this study, we conclude that input image resolution has a clear and consistent impact on the accuracy of age estimation in both DeepFace and InsightFace. Both frameworks achieve optimal performance at 224x224 pixels, with an MAE of 10.83 years (DeepFace) and 7.46 years (InsightFace). At low resolutions, MAE increases substantially, while very high resolutions also degrade accuracy. InsightFace is consistently faster than DeepFace across all resolutions.

Paper Structure

This paper contains 16 sections, 3 equations, 5 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background on Age Estimation
  4. Regression vs. Classification
  5. Age Estimation Processing Pipeline
  6. Datasets for Age Estimation
  7. Choice of Existing Frameworks
  8. Architectures for Age Estimation
  9. Dataset and Preprocessing
  10. Experimental Environment
  11. Experimental Design
  12. Evaluation Metrics
  13. Formulas Used
  14. Processing Time Measurement
  15. Results and Discussion
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Processing pipeline for age estimation rothe2016.
  • Figure 3: Example output from DeepFace: age, emotion, gender, and race predictions generated from a facial image serengil_benchmark_2024.
  • Figure 4: Example of a single image in seven different resolutions.
  • Figure 5: Screenshot of the Colab runtime (T4 GPU, Python 3).
  • Figure 6: Comparison of DeepFace and InsightFace across different image resolutions based on MAE, SD, and MedAE.